已入深夜,您辛苦了!由于当前在线用户较少,发布求助请尽量完整地填写文献信息,科研通机器人24小时在线,伴您度过漫漫科研夜!祝你早点完成任务,早点休息,好梦!

PRANCE: Joint Token-Optimization and Structural Channel-Pruning for Adaptive ViT Inference

计算机科学 人工智能 推论 接头(建筑物) 安全性令牌 修剪 频道(广播) 模式识别(心理学) 机器学习 工程类 计算机网络 农学 生物 建筑工程
作者
Ye Li,Chen Tang,Yuan Meng,Jiajun Fan,Zenghao Chai,Xinzhu Ma,Zhi Wang,Wenwu Zhu
出处
期刊:IEEE Transactions on Pattern Analysis and Machine Intelligence [IEEE Computer Society]
卷期号:48 (1): 283-298 被引量:19
标识
DOI:10.1109/tpami.2025.3605239
摘要

The troublesome model size and quadratic computational complexity associated with token quantity pose significant deployment challenges for Vision Transformers (ViTs) in practical applications. Despite recent advancements in model pruning and token reduction techniques speed up the inference speed of ViTs, these approaches either adopt a fixed sparsity ratio or overlook the meaningful interplay between architectural optimization and token selection. Consequently, this static and single-dimension compression often leads to pronounced accuracy degradation under aggressive compression rates, as they fail to fully explore redundancies across these two orthogonal dimensions. Therefore, we introduce PRANCE, a framework which can jointly optimize activated channels and tokens on a per-sample basis, aiming to accelerate ViTs' inference process from a unified data and architectural perspective. However, the joint framework poses challenges to both architectural and decision-making aspects. First, while ViTs inherently support variable-token inference, they do not facilitate dynamic computations for variable channels. To overcome this limitation, we propose a meta-network using weight-sharing techniques to support arbitrary channels of the Multi-Head Self-Attention (MHSA) and Multi-Layer Perceptron (MLP) layers, serving as a foundational model for architectural decision-making. Second, simultaneously optimizing the model structure and input data constitutes a combinatorial optimization problem with an extremely large decision space, reaching up to around $10^{14}$1014, making supervised learning infeasible. To this end, we design a lightweight selector employing Proximal Policy Optimization algorithm (PPO) for efficient decision-making. Furthermore, we introduce a novel "Result-to-Go" training mechanism that models ViTs' inference process as a Markov decision process, significantly reducing action space and mitigating delayed-reward issues during training. Additionally, our framework simultaneously supports different kinds of token optimization methods such as pruning, merging, and sequential pruning-merging strategies. Extensive experiments demonstrate the effectiveness of PRANCE in reducing FLOPs by approximately 50%, retaining only about 10% of tokens while achieving lossless Top-1 accuracy.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
今后的应助被好吃懒做采纳,获得10
1秒前
隐形的芹菜完成签到,获得积分10
1秒前
田様的应助被隐形的芹菜采纳,获得10
5秒前
pine发布了新的文献求助10
5秒前
hopen完成签到,获得积分10
6秒前
自然大侠完成签到,获得积分10
6秒前
科研通AI6.4的应助被Atlantis30采纳,获得10
7秒前
爱喝水完成签到 ,获得积分10
9秒前
9秒前
dolabmu完成签到 ,获得积分10
10秒前
斯文的清涟完成签到,获得积分20
12秒前
刘天歌完成签到 ,获得积分10
13秒前
一枚小豆完成签到,获得积分0
13秒前
大模型的应助被糖心采纳,获得10
14秒前
15秒前
15秒前
小豆包发布了新的文献求助10
15秒前
16秒前
SD512T发布了新的文献求助10
16秒前
我是老大的应助被狗头采纳,获得30
16秒前
dzy完成签到,获得积分10
16秒前
Jiaaa完成签到 ,获得积分10
17秒前
科研通AI6.3的应助被pero采纳,获得10
17秒前
难过的泥猴桃完成签到 ,获得积分10
17秒前
知性的藏鸟完成签到 ,获得积分10
17秒前
冰美式发布了新的文献求助10
18秒前
18秒前
18秒前
华无剑的应助被开心香岚采纳,获得10
20秒前
隐形曼青的应助被Atlantis30采纳,获得10
20秒前
勤奋的猫咪完成签到 ,获得积分10
20秒前
21秒前
PeterPan发布了新的文献求助10
22秒前
小怪兽完成签到 ,获得积分10
23秒前
听闻姑娘质比芝兰完成签到,获得积分10
24秒前
凌云客完成签到,获得积分10
25秒前
25秒前
milo完成签到 ,获得积分10
25秒前
25秒前
25秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
拉伸高分子熔体中缠结行为的模拟和小角中子散射研究 2000
Sex, Aging, and Death in a Medieval Medical Compendium: Trinity College Cambridge MS R.14.52, Its Texts, Language, and Scribe 500
高分子基功能界面水蒸发材料与应用研究 500
外场增强的水汽发电研究 500
基于结构调控的功能化水诱导发电材料与器件研究 500
基于界面调控的高效湿气发电材料与器件研究 500
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 计算机科学 工程类 有机化学 化学工程 生物化学 物理 复合材料 内科学 催化作用 光电子学 人工智能 细胞生物学 无机化学 物理化学 数学
热门帖子
关注 科研通微信公众号,转发送积分 8189908
求助须知:如何正确求助?哪些是违规求助? 9577068
关于积分的说明 21851352
捐赠科研通 7787862
什么是DOI,文献DOI怎么找? 3435817
关于科研通互助平台的介绍 2559861
邀请新用户注册赠送积分活动 2482718